Declarative Data-Driven Chart Animation

Animation is an effective means to help people understand the changes and connections between objects.In particular,when using visualization for data exploration and storytelling,the use of animation can make data more vivid and intuitive.Animation can not only help readers track the visual changes of the graphic element and establish a connection between states of the graphic element,but can also effectively attract and maintain people's attention.Therefore,visualization animation is widely used in major mainstream news media and visualization platforms(such as the New York Times,The Guardian,and FlowingData).The existing visualization animation creation methods mainly focused on key frames,templates or programming.However,these methods currently have many limitations,which make it hard to create the animation,as well as cumbersome operation and lack of comprehensibility.At the same time,the generated animation has low expressiveness and the consistency with the data perception cannot be guaranteed.Therefore,how to create an effective and customized visualization animation is one of the most challenging problems in today's visualization field.To address the above problems,this dissertation understands and explores the problem of visualization animation creation from the three levels of animation description,comprehensibility and interactive mode of animation creation.First of all,in order to enable users to adopt data-driven creative forms to create expressive animations with concise and precise descriptions,we characterize the design space of visualization animations into four parts(targets,divisions,timing and effects).We also propose a concise and full-featured declarative visualization animation programming language.After that,in order to improve the comprehensibility of animation and reduce the cost of interaction,we designed and implemented an interactive animation creation platform based on the expansion and optimization of the above programming language.The platform adopts a visual specification that can accurately describe the animation process and a data-driven automatic animation completion mechanism.Finally,in order to further improve and optimize the platform to have a more natural and intuitive interaction,we propose a multi-modal animation creation mode that supports sketching,gestures,voice and other manipulation methods.The main contributions of this dissertation are as follows:(1)A data-driven high-level programming language for visualization animationCanisIn order to describe animation concisely and accurately in a data-driven format,this dissertation proposes a concise declarative grammar,Canis.This grammar can help users quickly build an abstract relationship between data,visual encoding,and timing.Supports defining the grouping,time,effect and other parameters of the animation elements in a declarative manner,thereby creating a more expressive visualization animation.Canis specification will be compiled into an animation encoding format that supports efficient cross-platform renderiug.(2)An interactive authoring platform for chart animations that combines visual specification and automatic completion-CASTIn order to enhance the comprehensibility of animation and further reduce the construction cost,this dissertation designs a visual specification based on key frames to accurately describe the animation process.The specification uses visual components including key frames(groups),time bars,effect icons,etc.to represent the core parameters of the animation.Combined with the visual specification,this dissertation designed and implemented the interactive authoring system for chart animation,CAST.CAST uses optimized Canis for animation state management,and provides users with direct interaction with visual components to adjust animation statements.At the same time,the system uses a data-driven auto-completion mechanism to improve the efficiency and accuracy of key frames and key frame sequences.(3)A multi-modal authoring mode for chart animations-M2-CASTIn order to improve the intuitiveness and consistency of the animation creation process,and reduce the cognitive gap between the user's expression and execution of the target animation,this paper introduces multi-modal interaction into the system.This dissertation first designs and implements a state management mechanism for multimodal interactions,and gives full play to the advantages of the accuracy and directness of touch interactions and the freedom of expression of speech interactions to achieve complementary advantages.After that,a multi-modal animation authoring mode M2CAST was designed and implemented on this basis.The user can freely choose the interaction method to describe the expected animation during the creation process,without resorting to indirect interactive means such as mode switching and parameter calculation.The system will automatically identify the drawing content,instructions and other information,and automatically generate the corresponding animation content for the user to adjust and declare afterwards.